Post on 02-Aug-2018
Università degli Studi di Siena
DIPARTIMENTO DI ECONOMIA POLITICA
ALESSANDRO INNOCENTI MARIA GRAZIA PAZIENZA
Experimenter bias across gender differences
n. 438 - Ottobre 2004
ABSTRACT: This paper reports on a laboratory study that examined the influence of experimenter bias in the investment game. Specifically we explored the effect of changing the gender of the experimenter and compared it with the double blind treatment. Our findings show that the presence of a female experimenter influences reciprocity. We also provide further evidence on gender differences in trust and reciprocity. JEL CODES: C90, C91, J16. KEYWORDS: experimenter bias, gender differences, trust, reciprocity. The financial support of the University of Siena and the MIUR is gratefully acknowledged. We thank Francesco Lomagistro for valuable research assistance. Thanks also to the participants in the 2004 International ESA Meeting, Amsterdam, 11–14 June 2004, where a previous version of this paper was presented.
Alessandro Innocenti, Dipartimento di Economia Politica e LabSi, Laboratorio di Economia Sperimentale, Università degli Studi di Siena Maria Grazia Pazienza, Dipartimento di Studi sullo Stato, Università di Firenze
1. Introduction
An important issue in the methodology of laboratory techniques is experimenter
bias.1 This definition applies to laboratory environments in which experimenters’ acts
can unconsciously convey to subjects how they should behave in relation to some
characteristics of the design and consequently produce biased results. A different case is
participant bias, also known as demand characteristics, which applies to experiments in
which participants act in ways they believe correspond to what the experimenter is
looking for. Thus, if participants modify their spontaneous behaviour to match the real
or the presumed aims of the experimenter, results are also biased.
These sources of bias are differentiated by the fact that the former is explicitly
related to some specific act or characteristic of the experimenter, while the latter refers
generically to the experimental design, but it is not always easy to discriminate between
them. What matters more for experimental methodology is that in both cases some
features of the laboratory environment may induce subjects to change their choices in
order to comply with the experiment’s purpose. Once the experimenter becomes aware
of this effect, data interpretation has to be revised in order to assess the experimental
findings correctly.
The very robust finding that subjects prefer fair to maximized payoffs2 has been
attributed to the effect of experimenter’s observation. Hoffman, McCabe, Shachat and
Smith (1994) and Bolton and Zwick (1995) name this effect the anonymity hypothesis
and give two reasons for it. The first is related to subjects’ participation in future
experiments. If the experimenter’s presumed aim is to provide evidence against self-
seeking behaviour, a subject’s preference for fair payoffs could increase his or her
probability of being recruited again by the same experimenter. The second reason has a
more psychological flavour. Subjects could be concerned with the experimenter’s
judgement and believe that he or she disapproves of maximized choices.
1 It is only recently that experimental economists have become aware of this problem. In particular, Hoffman, McCabe and Smith (1996) argue that the subjects’ degree of social distance from the experimenter may affect subjects’ behaviour, especially determining expectations of reciprocity. Bolton and Zwick (1995) try to find evidence for the hypothesis that experimenter’s observation distorts subjects’ objectives by testing an ultimatum game. 2 Ledyard (1995) and Schram (2000) present extensive surveys documenting that, in a variety of experimental situations, subjects deviate from own-payoff-maximizing behaviour.
1
Both arguments can be criticized. Most experimenters prefer inexperienced subjects
to experienced subjects, in order to control better for learning processes. The beliefs of
experimenters about economic or moral principles are not easily predictable by the
participants in experiments. However, the importance of the experimental result of
subjects deviating from maximizing their own payoffs requires further analysis of this
issue.
In this light, Orne (1962) offers another plausible explanation: “The subject's
performance in an experiment might almost be conceptualized as problem-solving
behavior; that is, at some level he sees it as his task to ascertain the true purpose of the
experiment and respond in a manner which will support the hypotheses being tested.
Viewed in this light, the totality of cues which convey an experimental hypothesis to the
subject become significant determinants of subjects’ behavior.” (Orne 1962, p. 778).
Experimenter bias can become relevant because subjects have a real stake in
viewing their performance as meaningful. During the experiment, they are constantly
wondering about what the experimenter is trying to test and any cue they find useful to
answer this question can influence their behaviour. If the design incorporates explicit
suggestions as treatment variables, data interpretation is unbiased. Otherwise,
expectations about the experimental purpose can affect subjects’ behaviour and
consequently distort results.
Observable characteristics of the experimenter may be among these cues. Indeed
the object of the experimenter can also be inferred from his or her gender. For example,
the presence of a female experimenter in tests of gender differences may induce
participants to guess that the experiment’s purpose is related to discrimination against
women. This source of bias can also be reinforced if subjects are informed about the
gender of other participants.
Although several recent studies focus on gender effect,3 we are not aware of
laboratory experiments that try to analyse the effect of experimenter bias across gender
differences.4 Our study intends to provide evidence on this issue by testing the
3 See, among the others, the experimental studies on dictator games by Andreoni and Vesterlund (2001) and Eckel and Grossman (1998). 4 Andreas Ortmann and Lisa Tichy (1999) tried to deal with the experimenter bias problem by jointly conducting a test on gender differences in prisoner’s dilemma. However, the physical presence of two experimenters, one male and one female, does not remove the possibility that one of the two experimenters is perceived by the participants as the determinant one. For
2
differences between men’s and women’s individual choices in three different
treatments. The first and second treatments were conducted by a female and a male
experimenter respectively, while the third treatment adopted a double blind procedure.
We consider our experiment useful also because it offered additional evidence of gender
differences on trust and reciprocity, which is still a controversial issue in economics.
The object of our test was the investment game, also known as the trust game,
which was first tested by Berg, Dickhaut and McCabe (1995). Two players are paired
off anonymously and respectively named as the sender and the receiver. The sender is
given a certain amount of money and told that he or she can keep the entire amount or
send some or all of it to the receiver. Any money passed from the sender to the receiver
is tripled by the experimenter and then given to the receiver. The receiver can keep the
entire amount or give back some or all of it to the sender. When the sender receives the
amount sent back by the receiver the game ends. The backward induction solution of the
game predicts that the receiver will not send any money back. Anticipating the
receiver’s decision, the sender will not send any money to the receiver.
Results from earlier experiments are inconsistent with the conventional game
theory prediction. Table 1 offers a summary of the previous work on the investment
game.
Table 1. Experimental results on the investment game Berg, Dickhaut
and McCabe (1995)
Buchan, Croson and Johnson (2000)
Schwieren and Sutter (2003)
Burks, Carpenter and Verhoogen (2003)
Chaudhuri and Gangadharan (2003)
No history
Social history
Only U.S.
All Single role
Both roles, no
prior
Both roles, prior
Trust
51.6 53.6 49 67 65.7 65.0 65.2 47.3 43.3
Reciprocity
30.1 40.2 22 37 37.6 43.6 25.9 17.1 17.5
Trust = Average fraction sent (Amount sent / Initial endowment); Reciprocity = Average fraction returned (Amount sent back/ Amount received) Notes: In the social history treatment by Berg, Dickhaut and McCabe (1995) subjects were given a summary of the no history treatment results as part of the instructions. Buchan, Croson and Johnson’s (2000) international experiment tested the investment game in China, Japan, Korea and United States. Burks, Carpenter and Verhoogen (2003) compared three treatments: the first (single role) in which subjects played either sender or receiver role, the second (both roles, no prior) in which subjects played both roles but they did not know it before sending, and the third (both roles, prior) where subjects knew that they played both roles before any decisions were made.
example, the person who reads the instructions aloud is usually considered to lead the experiment.
3
Even if there are significant variations across tests, the relevance of altruistic
motivations and intentions to reciprocate was widely confirmed in the laboratory.5
Other studies examined gender differences in the investment game. Table 2
presents their results.
Table 2. Experimental results on gender differences in the investment game Croson and Buchan (1999) Chaudhuri and Gangadharan (2001) Men Women Men Women Trust 69.0 63.0 53.0 34.7 Reciprocity 28.6 37.4 14.7 19.8 Trust = Average fraction sent (Amount sent/Initial endowment); Reciprocity = Average fraction returned (Amount sent back/Amount received)
In both tests men exhibited greater levels of trust and lower levels of reciprocity
than women did, even if the difference between men and women was not statistically
significant for trust in Croson and Buchan (1999) and reciprocity in Chaudhuri and
Gangadharan (2001).
Our experiment adopted nearly the same design as those tests, including that each
subject played both roles of sender and of receiver, but we departed from them by
imposing the following variant: when participants played as senders, they were
informed about the gender of the receiver with whom they were paired off.
The relevance of gender pairing in bilateral relationships was proved
experimentally by Sutter, Bosman, Kocher and van Winden (2003). They found that
cooperation between players was lower when bargaining partners have the same gender
than when they have the opposite gender. In our design, the information about their
receiver’s gender aimed at focusing senders’ attention on the gender variable. In this
way, a clear signal about the experimental purpose was conveyed to all subjects. If the
experimenter’s presumed aim was to study the effect of gender differences on trust and
reciprocity, the experimenter’s gender might affect subjects’ behaviour. For example,
the effect of the presence of a female experimenter might be an increase in the degree of
trust in women or a change in women’s degree of trust in receivers of both sexes. More
5 Abbink, Irlenbusch and Renner (2000), Cox (2001) and Cox, Sadiraj and Sadiraj (2002) obtained similar results by testing the moonlighting game. In the moonlighting game, the sender can choose if she wants to give to the receiver part of her endowment or take up to half the endowment given to the receiver. The receiver can decide whether she wants to give or take money from the sender.
4
generally, we expected subjects in the double blind treatment to exhibit behaviour
significantly different from that in the other two treatments, because the signals
conveyed by the experimenter’s gender were absent. In addition, we expected subjects
to change their behaviour significantly in relation to the experimenter’s gender. We
might conjecture that the presence of the female experimenter could be perceived as
relatively unusual and consequently conveys a stronger signal than the presence of a
male experimenter about the experimental purpose.
To examine further the difference across treatments we chose not to inform
receivers about their sender’s gender. This feature of the design was intended to
differentiate factors influencing reciprocity and trust. If subjects, when they played as
receivers, did not know the gender of their paired sender, their decision to reciprocate
could be considered dependent only on the amount of money received and on the
experiment’s perceived purpose.
To summarize, our experiment intended to verify the effect of experimenter bias by
testing two hypotheses. The first hypothesis was that subjects’ behaviour in the double
blind treatment is significantly different from that in the other two treatments. The
second hypothesis was that subjects change their behaviour significantly in relation to
experimenter’s gender, presumably by increasing the degree of trust and reciprocity in
the female experimenter treatment.
The rest of the paper is organized as follows. The experimental design is introduced
in Section 2. In Section 3, we report the experimental results, focusing on the effect of
experimenter’s gender, sender’s gender and receiver’s gender. Section 4 offers some
concluding remarks.
2. Experimental design
The experiments were carried out in the spring of 2004. We submitted the
investment game to 94 subjects: 46 women and 48 men. They were undergraduate
students in economics from the University of Siena and students in political sciences
from the University of Florence, recruited from economics courses through billboards
posted around the campuses of the two universities.
5
The experiments were run manually. The participants were paid according to the
euros earned. There was no participation fee.
We ran three treatments. The only difference between the first and the second
treatment was that a female and a male respectively played the role of experimenter.6 It
was made clear to subjects that in these treatments only the experimenter was able to
attribute individual choices to individual people. However, anonymity between subjects
was guaranteed. The third control treatment adopted a double blind procedure.
Table 3 presents the number of participants for each session and treatment.
Table 3. Number of participants per treatment by gender Session Treatment Participants (female + male)
1 Female experimenter 8 + 6 2 Female experimenter 7 + 9 3 Male experimenter 7 + 9 4 Male experimenter 8 + 8 5 Double blind 8 + 8 6 Double blind 8 + 8 46 + 48
In the female and male treatments, subjects were first identified by numbers. These
numbers were randomly assigned and determined the pairings of senders and receivers.
Then each subject was directed to an isolated desk to make his or her decision privately.
There they received written instructions. The first part of the instructions was read aloud
by the experimenter of the appropriate gender. The second part contained a short
questionnaire that was answered at the end of the experiment.
When the experiment began, subjects were given a large unmarked envelope which
contained the money to be invested (5 euros which could be transferred in steps of half
units), a card marked with the identification number and a small envelope marked with
a circle that was either pink or blue. Subjects were asked to remember their numbers.
The correspondence between each number and each participant remained unknown to
the other participants but not to the experimenter and this was made clear to the
participants. Subjects were also informed that if the circle was pink (blue) the person to
whom they were to send money was a female (male). In this way, the sender knew the
receiver’s gender but the sender’s gender remained unknown to the receiver.
6
Senders decided how many euros to keep and how many euros they wish to send to
their partner by inserting them in the small envelope. The experimenter collected the
small envelopes, recorded the amount sent, tripled it and placed the tripled money into
the same envelope for delivery to the appropriate receiver. Receivers then opened their
envelopes and decided how much of the money received to return to the sender. The
experimenter again collected the envelopes, recorded the amounts returned and gave the
envelopes back to the senders. Subjects were informed in the written instructions that
they would be playing both roles of sender and receiver but also that the receiver to
whom they were paired as sender, would not be their sender when they played the role
of receiver.
In the double blind treatment, the design had to guarantee to the participants that
the experimenter was unable to attribute individual choices to individual subjects and to
avoid any hint about the experimenters’ gender. Subjects were gathered in a room where
two undergraduate students, a male and a female previously instructed to play the role
of monitors, gave them instructions to read privately. When the experiment began,
subjects were given a large unmarked envelope, which contained 5 euros, a smaller
envelope and a numbered identification card. The small envelope was marked with a
pink or a blue circle. As in the previous treatments, subjects were informed that the
coloured circle on the small envelope identified the gender of their paired receiver.
Moreover, they were asked to remember their numbers. The correspondence between
these numbers and identities of the subjects remained unknown to the experimenters, to
the monitors and to the other participants at all times and this was made clear to the
participants.
Once senders had decided how much money to send to their partners in the small
envelopes, they had to insert the identification cards in the smaller envelopes.
Moreover, each subject had to write on the identification card the letter F, if female, or
the letter M, if male. The sealed envelopes were collected in a closed urn and were
taken by the monitor to the experimenters in another room. After recording the amount
sent and tripling it, the experimenters marked each larger envelope with the number
identifying a receiver of the appropriate gender. The envelopes in the closed urn were
6 The authors served as experimenters for the two treatments. In each session, there was also an assistant of the same gender as the experimenter.
7
delivered by the monitor to the subjects’ room. At this time, subjects were called one at
a time by the monitor. Once called, a subject had to privately choose the envelope with
her or his identification number from the urn placed on an isolated desk. Having
decided how much of the money received to return to the sender, subjects sealed their
envelopes. The monitors again collected the envelopes, and took them to the
experimenters’ room, where the experimenters recorded the amounts returned and gave
the envelopes back for distribution to senders by the same procedure used before. When
the experiment was over, all subjects left the room without revealing their identities.
To summarize, our main variations with respect to the reference design by Berg,
Dickhaut and McCabe (1995) were the following ones:
1. All subjects played both roles (sender and receiver) and they knew it before any
decisions were made.7
2. Senders were informed of their receiver’s gender, but receivers did not know
their sender’s gender.
3. Only the third treatment was double blind, while in the first and in the second
treatment, the experimenter was able to attribute individual choices to individual
subjects and this was made clear to the participants.
3. Results
Our discussion of the experimental findings addresses the question of whether
men and women made different choices across the three treatments, first for trust, and
then for reciprocity.
Table 4 presents senders’ behaviour in the three treatments across gender.
7 Burks, Carpenter and Verhoogen (2003) made the same assumption in their third treatment, which was named as ‘both roles, prior’ (see Table 1).
8
Table 4. Trust per treatment by sender’s gender
Treatment Female experimenter
Male experimenter
Double blind Total
Sender’s gender Mean StdD Mean StdD Mean StdD Mean StdD
Women 43.3 35.4 39.3 22.2 32.5 23.8 38.3 27.4
Men 28.0 27.0 24.1 18.4 39.4 27.2 30.4 24.8
Total 35.7 31.9 31.3 21.4 35.9 25.4 34.3 26.3
Trust = Average fraction sent (Amount sent/Initial endowment)
The trust indicator is computed as the average ratio between the amount sent and
the initial endowment. Contrary to the previous experimental evidence (see Table 2),
women showed on average a higher degree of trust than men, but the difference
between the two means (38.3 and 30.4) was not significantly different from zero. In the
aggregate, trust (34.3) was lower than in the previous tests presented in Table 1.
However, the results of the double blind treatment were very similar to that of
Chaudhuri and Gangadharan (2001), in particular in the average fraction sent by women
(32.5 versus 34.7).
We found similar values of trust across the three treatments (35.7, 31.3, and 35.9).
The differences between women and men were not significant as well. Although women
trusted less in the double blind treatment (32.5), and men trusted less in the male
experimenter treatment (24.1), the difference in the fraction of the amount sent between
women and men was not significant using either a t-test (with a t-value of 0.04 and a p-
value of 0.97) or a Wilcoxon rank sum test (with a z-value of 911 and a p-value of
0.63).
In Figure 1, data on trust are summarized with box plots. The plots report the
median of the data distribution, and the interquartile range to measure the data
dispersion. The dispersion of data, which is higher in the female experimenter
treatment, explains why the mean differences are not statistically significant.
Nevertheless, the figure shows that the difference between men’s and women’s median
values was greater than the difference between the corresponding average values.
9
Figure 1. Distribution of trust per treatment and sender’s gender
Women Men
Sender's Gender
0,00
20,00
40,00
60,00
80,00
100,00
Trus
t (Ex
pect
atio
n of
reci
proc
ity) 94
59Experimenter's
GenderDouble blindFemaleMale
In our design, senders were informed of their paired receiver’s gender. This
information was specifically given to emphasize the effect of experimenter bias. Table 5
presents the average values of trust by receiver’s gender. In the aggregate, senders
trusted more men (35 %) than women (33%), but the difference between the two
average values was not significant and the medians were nearly identical (Figure 2). In
the male and in the female experimenter treatments, senders exhibited a higher degree
of trust in men than in women, while in the double blind treatment the reverse was true.
There was no evident gender pairing effect because the degree of trust was quite similar
across pairs of the same gender and of opposite gender.
10
Table 5. Trust per treatment by receiver’s gender Treatment
Female
experimenter Male
experimenter Double blind
Table total
Total (N=94) Women 30.67 26.67 41.88 33.26 Men 40.67 35.29 30.00 35.21 Receiver's
gender Total 35.67 31.25 35.94 34.26 Male senders (N=48)
Women 21.25 21.00 50.00 30.00 Men 35.71 28.57 28.75 30.91 Receiver's
gender Total 28.00 24.12 39.38 30.42 Female senders (N=46)
Women 41.43 38.00 33.75 37.50 Men 45.00 40.00 31.25 38.85 Receiver's
gender Total 43.33 39.33 32.50 38.26 Trust = Average fraction sent (Amount sent/Initial endowment)
The box plots depicted in Figure 2 show that the amount of money sent to both men
and women was more dispersed in the female experimenter treatment than in the other
two treatments.
Figure 2. Distribution of trust per treatment and receiver’s gender
Double blind Female Male
Experimenter's Gender
0,00
20,00
40,00
60,00
80,00
100,00
Trus
t (Ex
pect
atio
n of
recip
rocit
y) 94
59Receiver's Gender
WomenMen
11
Apart from this distinction, our findings show that there was no systematic
difference in trust indicator among treatments. Experimenter gender did not seem to
significantly influence senders’ choices.
Next, we turn to the analysis of receivers’ choices. Table 6 presents an indicator of
reciprocity, i.e. the average fractions returned by receivers.
Table 6. Reciprocity per treatment by receiver’s gender
Treatment Female experimenter
Male experimenter
Double blind Total
Receiver’s gender Mean StdD Mean StdD Mean StdD Mean StdD
Women 27.3 27.0 21.0 17.9 20.4 20.8 22.6 21.5
Men 30.9 22.6 9.7 17.7 18.9 19.4 19.0 21.1
Total 29.1 24.4 15.2 18.4 19.7 19.8 20.8 21.3
Reciprocity = Average fraction returned (Amount sent back/Amount received) The stronger pattern of behaviour was that the presence of a female experimenter
induced subjects of both genders to reciprocate more than in the other treatments. This
result is also shown in Figure 3,8 which presents the median values and the interquartile
range of the distribution in the three treatments.
8 The choice of the proportion of the amount sent back by each receiver as the indicator of reciprocity leads to some missing cases. If the receiver was sent nothing by the paired sender, he or she is excluded from computation as a missing variable. Otherwise, if the computed value is zero this means that the receiver gives back nothing of what he receives.
12
Figure 3. Distribution of the average values of reciprocity per treatment
The median values in the male experimenter treatment and in the double blind
treatment were quite similar and significantly lower than in the female experimenter
treatment. Statistical tests also corroborate this difference for the average values. Table
7 shows that both the t-test and the Wilcoxon test were significant at 95%.
Double blind Female Male
Experimenter's Gender
0,00
10,00
20,00
30,00
40,00
50,00
60,00
70,00
Rec
ipro
city
Table 7. Statistical tests for reciprocity among treatments
t test for equality of means N Mean Mean diff. T Sig. (2-tailed)Female experimenter treatment 24 29.12 Male experimenter and double blind treatments 59 17.44 –11.67 –2.32 0.02
Wilcoxon test Wilcoxon W 2289.5 Z –1.95 Sig. (2-tailed) 0.052
The box plots in Figure 4 show how male receivers particularly showed a lower
degree of reciprocity in the male experimenter treatment. This effect is reinforced by the
fact that the median value is zero.
13
Figure 4. Distribution of reciprocity per treatment by subject’s gender
The propensity to be “fairer” in the female experimenter treatment is also
confirmed by the statistical analysis of men’s behavior.9 Table 8 shows that the presence
of a female experimenter induced male subjects to reciprocate significantly more than in
the other two treatments, using either a t-test or a Wilcoxon rank sum test.
Women Men
Subject's Gender
0,00
10,00
20,00
30,00
40,00
50,00
60,00
70,00R
ecip
roci
ty
57
51
Experimenter's GenderDouble blindFemaleMale
Table 8. Statistical tests for reciprocity among treatments (men only)
t test for equality of means N Mean Mean diff. T Sig. (2-tailed)Female experimenter Treatment 12 30.92 Male experimenter and double blind treatments 30 14.30 –16.62 –2.44 0.02
Wilcoxon test Wilcoxon W 563.5 Z –2.35 Sig. (2-tailed) 0.02
9 The amount of money received from the paired sender and the percentage of money sent back to the paired sender was positively correlated with the Spearman’s correlation coefficient of 0.46 (0.576 for women and 0.372 for male), which was significative at the 0.01 level in all cases.
14
These outcomes may be interpreted as evidence of experimental bias for
reciprocity. The presence of the female experimenter could have influenced subjects’
expectations and changed their spontaneous behaviour. According to our proposed
justification for experimenter bias, receivers increased the degree of reciprocity to fulfil
the perceived experimental purpose that was presumably the testing of trustworthiness.
But experimenter bias would have been effective only for the treatment in which the
signal used to infer this purpose was perceived by subjects as more evident, specifically
the presence of a female experimenter. A rationale for this outcome is that the subjects
perceived the female experimenter as unusual. The effect was to increase the degree of
reciprocity behaviour in all subjects, male and female. The same effect did not emerge
for trust. An explanation may be found in the difference between the factors influencing
trust and reciprocity, as we discuss in the next section, which draws some conclusions.
4. Conclusions
Our experiment intended to test the effect of experimenter bias in the laboratory.
Among the possible features of an experimental environment eligible for this purpose,
experimenter’s gender was a straightforward option, especially if the design adopts
gender differences as a treatment variable. Moreover, we chose to test the investment
game that has been widely investigated in the last decade. Another reason to opt for the
investment game was that two different kinds of behaviour could be studied in the same
experiment, i.e. trust and reciprocity. Even in the one-shot version of the game, these
two tendencies can be the product of different motivational traits. Trust can be
perceived as an investment and therefore be dependent mainly on risk aversion;
reciprocity or trustworthiness is usually more affected by psychological or ethical
factors, which range from the perceived kindness of trusting to inequality aversion. This
distinction motivated us in defining both theoretical premises and experimental design.
The theoretical starting point was the conjecture that experimental subjects try to
discover the true purpose of the experiment and may change their spontaneous
behaviour to confirm the experimenter’s expectations. Any signal useful to perform this
task can influence their behaviour. The gender of the experimenter may be the source of
15
a bias if gender differences are explicitly considered as variables determining choices.
The experimental design took into account the distinction between trust and reciprocity
by differentiating subjects’ information: trusters were informed of their paired trustee’s
gender but trustees did not know their paired truster’s gender. We observed behaviour in
three different treatments: the female experimenter treatment and the male experimenter
treatment, conducted respectively by the female and male authors of this paper, and the
double blind treatment, where complete anonymity between subjects and between
subjects and experimenter was guaranteed. The proposed interpretation of experimenter
bias led to the prediction that subjects would modify their behaviour in relation to the
experimenter’s gender and to the way it is related to the experimental purpose. The
analysis of the data indicates that (i) there was no significant evidence that
experimenter’s gender influenced the degree of trust; (ii) there was statistically relevant
evidence that the presence of a female experimenter induced a higher degree of
reciprocity than in other treatments; (iii) this effect concerned all subjects, regardless of
their gender.
We interpret these findings as generally supportive of the relevance of experimenter
bias. If the cues provided by laboratory environment are clear enough to convey a
specific experimental purpose, as the presence of a female experimenter in a test of
gender effect did, subjects’ behaviour is significantly affected. In our experiment, trust
was unbiased presumably because it was perceived as a behavioural pattern dependent
on risk attitude and self-seeking behaviour. On the contrary, the tendency to reciprocate
was perceived as dependent on considerations of fairness that were associated by the
subjects with the presence of the female experimenter.
Finally, our study provides new evidence on gender differences in trust and
reciprocity. The degree of trust and reciprocity was generally lower than in previous
experiments. Women trust and reciprocate more than men, but while the result for reci-
procity is in keeping with previous studies that for trust is not. We also found that gender
differences were higher for trust than for reciprocity, although not statistically
significant. Our results corroborate the view that gender effect depends more on the
details of the experimental structure than on behavioural differences among men and
women.
16
A number of questions are left unanswered by our study. First, is the bias produced
by the presence of the female experimenter generally true or does it vary between
countries? Second, how is it possible to manipulate an experimental environment to
nullify gender effect in the double blind treatment? Third, are experiments on gender
effect necessarily biased when subjects are informed of other subjects’ genders? We
consider these issues relevant for the study of gender differences in the laboratory and
we intend to address them in future experiments.
References
Abbink, Klaus, Bernd Irlenbusch and Elke Renner, (2000), “The Moonlighting Game: An
Empirical Study on Reciprocity and Retribution.” Journal of Economic Behavior and
Organization, 42, pp. 265–77.
Andreoni, James, and Lisa Vesterlund, (2001), “Which is the Fairer Sex? Gender Differences in
Altruism.” Quarterly Journal of Economics, 115, pp. 293–311.
Bacharach, Michael, Gerardo Guerra and Daniel Zizzo, (2001), “Is Trust Self-Fulfilling? An
Experimental Study.” Discussion paper, Department of Economics, University of
Oxford.
Berg, Joyce, John Dickhaut and Kevin McCabe, (1995), “Trust, Reciprocity, and Social
History.” Games and Economic Behavior, 10, pp. 122–42.
Bolton, Gary E. and Rami Zwick, (1995), “Anonymity versus Punishment in Ultimatum
Bargaining.” Games and Economic Behavior, 10, pp. 95–121.
Buchan, Nancy R., Rachel T.A. Croson and Eric J. Johnson, (2000), “Trust and Reciprocity: An
International Experiment.” Discussion paper, University of Wisconsin-Madison,
University of Pennsylvania and Columbia University.
Burks, Steven, Jeffrey Carpenter and Eric Verhoogen, (2003), “Playing both roles in the trust
game.” Journal of Economic Behavior and Organization, 51, pp. 195–216.
Chaudhuri, Ananish and Lata Gangadharan, (2001), “Gender Differences in Trust and
Reciprocity.” Discussion paper, Wellesley College and University of Melbourne.
Chaudhuri, Ananish and Lata Gangadharan, (2003), “Sending Money in the Trust Game: Trust
or Other Regarding Preferences.” Discussion paper, Wellesley College and University
of Melbourne.
17
Cox, James C., (2001), “On the Economics of Reciprocity.” Discussion paper, University of
Arizona.
Cox, James C., Klarita Sadiraj and Vjollca Sadiraj, (2002), “Trust, Fear, Reciprocity, and
Altruism.” Discussion paper, University of Arizona and University of Amsterdam.
Croson, Rachel, and Nancy Buchan, (1999), “Gender and Culture: International Experimental
Evidence from Trust Games.” American Economic Review, 89, pp. 386–91.
Eckel, Catherine C., and Philip J. Grossman, (1998), “Are Women Less Selfish than Men:
Evidence from Dictator Experiments.” The Economic Journal, 108, pp. 726–35.
Hoffman, Elizabeth, Kevin McCabe, Keith Shachat and Vernon L. Smith, (1994), “Preferences,
Property Rights, and Anonymity in Bargaining Games.” Games and Economic
Behavior, 7, pp. 346–80.
Hoffman, Elizabeth, Kevin McCabe and Vernon L. Smith, (1996), “Social Distance and Other-
Regarding Behavior in Dictator Games.” American Economic Review, 86, pp. 653–60.
Ledyard, John O., (1995), “Public goods: a survey of experimental research”, in The Handbook
of Experimental Economics, edited by J. Kagel and A. Roth, Princeton: Princeton
University Press.
Orne, Martin T., (1962), “On the social psychology of the psychological experiment: with
particular reference to demand characteristics and their implications.” American
Psychologist, 17, pp. 776–83.
Ortmann, Andreas, and Lisa K. Tichy, (1999), “Gender differences in the laboratory: evidence
from prisoner’s dilemma games.” Journal of Economic Behavior and Organization, 39,
pp. 327–39.
Schram, Arthur, (2000), “Sorting Out the Seeking: Rents and Individual Motivation.” Public
Choice, 103, pp. 231–58.
Schwieren, Christiane, and Matthias Sutter, (2003), “Trust in cooperation or ability? An
experimental study on gender differences.” Discussion paper, University of Innsbruck.
Sutter, Matthias, Ronald Bosman, Martin Kocher and Frans van Winden, (2003), “Experimental
evidence of the importance of gender pairing in bargaining.” Discussion paper, Max
Planck Institute, University of Innsbruck, De Nederlandsche Bank and CREED/
University of Amsterdam.
18
Appendix Instructions for all participants A. FEMALE AND MALE EXPERIMENTER TREATMENTS This is an experiment in the economics of decision-making. The Ministry of University
and the University of Siena have provided funds to conduct this research. The instructions you are about to read are self-explanatory. If you follow them closely and make appropriate decisions, you can earn an amount of money that will be given to you in cash at the end of the experiment. If you have any questions, please raise your hand and the experimenter will come to you and answer your question.
In this experiment, each of you will be paired with a different person. You will not be told who these people are either during or after the experiment nor will they be told who the others are. The only information you will have is the gender of the person to which you are paired. The experimenter will be in charge of the envelopes as explained below. In addition, he will verify that the instructions have been followed as they appear here.
Each person will be given 5 euros as a show-up fee for this experiment. Each person will have the opportunity to send in an envelope, some, all or none of their show-up fee to the person which you are paired to. The amount sent will be tripled. For example, if you send an envelope that contains 2 euros, the envelope will contain 6 euros when it reaches the paired person. If you send an envelope that contains 4 euros, the envelope will contain 12 euros when it reaches the paired person. The paired person will then decide how much money to send back to you and how much money to keep.
Each person will play both roles in the experiment. Each of you will be paired with two people. In one pair, you will be the person who decides how many of the 5-euro show-up fee to send to another person, who receives the amount sent tripled by the experimenter. In the other pair, you will be the person who receives the amount sent by another person and tripled by the experimenter and you will decide to send back some, all, or none of the amount received to the sender. So each of you will take two decisions. However, the important thing to bear in mind is that you are not paired with the same person as sender and receiver. Rather you are paired with two different people.
The remainder of these instructions will explain exactly how the experiment is run. This experiment is structured so that no one, except the experimenter, will know the personal decision of people. Since your decision is private, we ask that you do not tell anyone your decision during, or after, the experiment.
The experiment is conducted as follows: a number of large unmarked envelopes have been placed in a box. Each of these envelopes contains 5 euros as a show-up fee for this experiment, a card marked with an identification number that you are asked to remember, and a smaller envelope marked with a circle, which will be coloured pink or blue. The experimenter will hand one person at a time an unmarked envelope from the box. Once a person has an envelope, he or she will privately open the unmarked envelope and place as many euros in the smaller circled envelope as they want, keeping the rest. Examples: (1) put 2 euros in the smaller envelope and keep 3 euros; (2) put 4 euros in the smaller envelope and keep 1 euro. These are examples only; the actual decision is up to each person.
It is important to keep in mind that the person who received the amount you sent will be a female if the smaller envelope is marked with a pink circle and he will be a male if the smaller envelope is marked with a blue circle. This process will continue until everyone has made his or her decision.
Once a person has made a decision, the experimenter will collect all the larger, unmarked envelopes, and return them to the box. Notice that each returned envelope will look exactly the same.
19
After all the envelopes have been put in the return box, the experimenter will then privately, one at a time, take the smaller envelopes out of the larger envelopes, record on a sheet of paper the number of the identification card and the amount of money inside the smaller envelope. The experimenter will then triple the amount of money in the smaller envelope and place the smaller envelope back into the larger envelope. At this point, the experimenter will transfer the envelopes in the return box.
The experimenter will then give to each person, one at a time, an unmarked envelope from the box. Each of you will privately open the larger envelope and must decide how many euros to leave in the smaller envelope. The person keeps the remaining euros. The smaller envelope should then be placed in the larger envelope. When everyone has had the opportunity to make his or her decision, the experimenter will collect the larger envelopes and return them to the box. The experimenter will then privately, one at a time, open the larger envelopes and record how much is in the smaller envelope. After recording how much was in the smaller envelope, the experimenter will put the smaller envelope in the larger envelope, and will place them back in the return box.
Then experimenter will choose one person at a time to go to the box marked return envelopes to retrieve the smaller envelope with the appropriate identification number marked on it. Do not open your envelope yet. This process will continue until everyone has retrieved his or her envelope and returned to his or her seat. When everyone is finished, experimenter will ask if everyone has retrieved the correct envelope. If the people have all taken the correct envelope, then the experiment is finished. If, however, an envelope has ended up with the wrong person, then the experimenter will collect all the smaller envelopes again and the process will repeat until everyone has the correct envelope.
Before leaving the room you, everyone will be asked to fill out a short questionnaire. At the top of the questionnaire, you will be asked for the card identification number. Please do not forget to include this information. Once you have finished the questionnaire, you will be asked to put it in the box placed at the back of the room.
Please raise your hand if you have any questions regarding how the experiment will proceed.
B. DOUBLE BLIND TREATMENT This is an experiment in the economics of decision-making. The Ministry of University
and the University of Siena have provided funds to conduct this research. The instructions you are about to read are self-explanatory. Two of us have been chosen as monitors and will check that the instructions have been followed as they appear here. However, they will not answer any questions during this experiment. If you have any doubts, you should read back through these instructions. Now that the experiment has begun, we ask that you do not talk, at all, during this experiment. If you follow these instructions closely and make appropriate decisions, you can earn an amount of money that will be given to you in cash at the end of the experiment.
In this experiment, each of you will be paired with a different person. You will not be told who these people are either during or after the experiment nor will they be told who the others are. The only information you will have is the gender of the person to which you are paired.
Each person will be given 5 euros as a show-up fee for this experiment. Each person will have the opportunity to send in an envelope, some, all or none of their show-up fee to the person which you are paired to. The amount sent will be tripled. For example, if you send an envelope that contains 2 euros, the envelope will contain 6 euros when it reaches the paired person. If you send an envelope that contains 4 euros, the envelope will contain 12 euros when it reaches the paired person. The paired person will then decide how much money to send back to you and how much money to keep.
Each person will play both roles in the experiment. Each of you will be paired with two people. In one pair, you will be the person who decides how much of the 5-euro show-up fee to
20
send to another person, who receives the amount sent tripled by the monitors. In the other pair, you will be the person who receives the amount sent by another person and tripled by the experimenter and you will decide to send back some, all, or none of the amount received to the sender. So each of you will take two decisions. However, the important thing to bear in mind is that you are not paired with the same person as sender and receiver. Rather you are paired with two different people.
The remainder of these instructions will explain exactly how the experiment is run. This experiment is structured so that no one, including the experimenters and the monitors, will know the personal decision of people. Since your decision is absolutely private, we ask that you do not tell anyone your decision during, or after, the experiment.
The experiment is conducted as follows: a number of large unmarked envelopes have been placed in a box. Each of these envelopes contains 5 euros as a show-up fee for this experiment, a card marked with an identification number that you are asked to remember, and a smaller envelope marked with a circle, which will be coloured pink or blue. Then monitors will call one person at a time to go to the isolated box placed in the front of the room. Each person will take an unmarked envelope from the box and will come back to his or her isolated desk.
Once a person has an envelope, he or she will privately open the unmarked envelope and wrote on the identification card the letter F if he is a female or the letter M if he is a male. Please do not forget to include this information. Then each person place as many euros in the smaller circled envelope as they want, keeping the rest. Examples: (1) put 2 euros in the smaller envelope and keep 3 euros; (2) put 4 euros in the smaller envelope and keep 1 euro. These are examples only; the actual decision is up to each person. It is important to keep in mind that the person who received the amount you sent will be a female if the smaller envelope is marked with a pink circle and a male if the smaller envelope is marked with a blue circle. This process will continue until everyone has made his or her decision.
Once a person has made a decision, he or she will put the smaller envelope, and the identification card in the larger envelope. Then the monitors will call one person at a time to go to the isolated box. Each person will put the larger envelope into the box. Notice that each returned envelope will look exactly the same and neither monitors nor others will be able to attribute individual choices to individual subjects.
After all the envelopes have been put in the return box, the monitors will then privately, one at a time, take the smaller envelopes out of the larger envelopes, record on a sheet of paper the letter and the number written on the identification card and the amount of money inside the smaller envelope. The monitors will then triple the amount of money in the smaller envelope, place the smaller envelope back into the larger envelope, and write an identification number on the larger envelope. At this point, the monitors will transfer the envelopes in the return box.
The monitors will then call one person at a time to go to the isolated box to retrieve the larger envelope with his or her identification number marked on it. Do not open your envelope yet. This process will continue until everyone has retrieved his or her appropriate envelope and returned to his or her seat. When everyone is finished, monitors will ask if everyone has retrieved the correct envelope. If the people have all taken the correct envelope, then the experiment will continue. If, however, an envelope has ended up with the wrong person, then the monitors will call one person at a time again and the process will repeat until everyone has the correct envelope.
Then each of you will privately open the larger envelope and must decide how many euros to leave in the smaller envelope. The person keeps the remaining euros. The smaller envelope should then be placed again in the larger envelope. When everyone has had the opportunity to make his or her decision, the monitors will call again one person at a time. Each person will return the larger envelopes to the box. The monitors will then privately, one at a time, open the larger envelopes and record how much is in the smaller envelope. After recording how much was in the smaller envelope, the monitors will put the smaller envelope in the larger envelope, and will place them back in the return box.
21
Then monitors will call one person at a time to go to the box marked return envelopes to retrieve the smaller envelope with the appropriate identification number marked on it. Do not open your envelope yet. This process will continue until everyone has retrieved his or her envelope and returned to his or her seat. When everyone is finished, monitors will ask if everyone has retrieved the correct envelope. If the people have all taken the correct envelope, then the experiment is finished. If, however, an envelope has ended up with the wrong person, then the monitors will collect all the smaller envelopes again and the process will repeat until everyone has the correct envelope.
At this time, you should take all your belongings and leave the building when you are done. When everyone in the room has left, the experiment is over, and the monitors will be paid for their participation.
22